Hello. So, I was testing Deno 1.9.2 with a workload that involves Temporal and a Postgres database. I noticed that after a few minutes, the 2GB VPS just ran out of memory after some batches of processing. I did the following:

1. Tried to profile using --inspect, but I noticed that it never went above 20MB, so it was probably not a "normal" JavaScript leak. I think Deno lacks GC information in the DevTools profiler, so that wasn't helpful.
2. I proceeded to compile Deno in debug mode, upload to VPS, run with the workload using LD_PRELOAD=/usr/lib/libtcmalloc.so and a HEAPPROFILE.
3. The results were in. And looks like it's Intl.DateTimeFormat that is at fault.
   
4. Then I looked for similar issues. Found: Memory leak when using Intl.DateTimeFormat.format() and Intl.NumberFormat.format() nodejs/node#31914, seems related, and listed as fixed for v8, but let's try the repro. Created a file leak.ts, ran with deno and watched the resident memory climb up to 5.4G. Ran with node, it also looks "leaky", but it only goes up to 1.3G.

setInterval(() => {
  var a;
  for (let i = 0; i < 100; i++) {
    a = new Intl.DateTimeFormat("en").format(Date.now());
  }
}, 1);

5. The same code with a similar object, Intl.NumberFormat, peaks at 100M for node, and 258M for deno. I assume that is normal.

setInterval(() => {
  var a;
  for (let i = 0; i < 100; i++) {
    a = new Intl.NumberFormat("en").format(3);
  }
}, 1);

6. I couldn't change this memory ceiling by using --v8-flags=--max-old-space-size, so I guess it's completely external to that.

I think it's probably an issue in V8 itself that has not been correctly resolved, and deno and node both handle it differently. I will try to change Temporal to reuse the same Intl.DateTimeFormat instance, for now, but it would be nice if this could be fixed.